This invention relates to a contact device for making connection to an electronic circuit device and to a method of fabricating such a contact device.
An important aspect of manufacture of integrated circuit chips is the testing of the circuit embodied in the chip in order to verify that it operates according to specifications. Although the circuit could be tested after the chip has been packaged, the expense involved in dicing the wafer and packaging the individual chips makes it preferable to test the circuit as early as possible in the fabrication process, so that unnecessary efforts will not be expended on faulty devices. It is therefore desirable that the circuits be tested either immediately after wafer fabrication is completed, and before separation into dice, or after dicing, but before packaging. In either case, it is necessary to make electrical connection to all the circuits' external connection points (usually bonding pads) in a non-destructive way, so as not to interfere with subsequent packaging and connection operations.
U.S. Pat. No. 5,221,895 discloses a probe for testing integrated circuits. The probe includes a stiff metal substrate made of beryllium copper alloy, for example. The substrate is generally triangular in form and has two edges that converge from a support area toward a generally rectangular tip area. There is a layer of polyimide over one main face of the substrate, and gold conductor runs are formed over the layer of polyimide. The conductor runs and the metal substrate form microstrip transmission lines. The conductor runs extend parallel to one another over the tip area and fan out toward the support area. A contact bump is deposited on the end of each conductor run that is on the tip area. The tip area of the substrate is slit between each two adjacent conductor runs whereby the tip area is divided into multiple separately flexible fingers that project in cantilever fashion from the major portion of the substrate.
The probe shown in U.S. Pat. No. 5,221,895, is designed to be used in a test station. Such a test station may include four probes having the configuration shown in U.S. Pat. No. 5,221,895, the probes being arranged in an approximately horizontal orientation with their contact bumps facing downwards, with the four rows of contact bumps along four edges of a rectangle. The DUT is generally rectangular and has connection pads along the edges of one face. The DUT is placed in a vacuum chuck with its connection pads upwards. The vacuum chuck drives the DUT upward into contact with the probe, and overdrives the DUT by a predetermined distance from first contact. According to current industry standards, such a test station is designed to produce a nominal contact force of 10 grams at each connection pad. Therefore, the amount of the overdrive is calculated to be such that if contact is made at all connection pads simultaneously, so that each contact bump is deflected by the same amount, the total contact force will be 10 grams force multiplied by the number of connection pads.
If the material of the probe substrate is a beryllium copper alloy and each flexible finger has a length of about 0.75 mm, a width of about 62 .mu.m, and a height of about 250 .mu.m, and the probe is supported so that the mechanical ground is at the root of the fingers, the contact force produced at the tip of the finger is about 7.7 grams for each micrometer of deflection of the tip of the finger. Therefore, if the contact bumps at the tips of the fingers are coplanar and the connection pads of the DUT are coplanar, and the plane of the contact bumps is parallel to the plane of the connection pads, an overdrive of about 1.3 .mu.m from first contact will result in the desired contact force of 10 grams at each connection pad. However, if one of the connection pads should be 1.3 .mu.m farther from the plane of the contact bumps than the other connection pads, when the DUT is displaced by 1.3 .mu.m from first contact, there will be no contact force between this connection pad and its contact bump, and all the contact force that is generated will be consumed by the other contacts. If one assumes that the contact force at a connection pad must be at least 50 percent of the nominal contact force in order for there to be a reliable connection, then the maximum variance from the nominal height that this design will accommodate is .+-.0.7 .mu.m. However, the height variations of contact bumps and connection pads produced by the standard processes currently employed in the semiconductor industry exceed 5 .mu.m. Furthermore, even if the contact bumps are coplanar and the connection pads are coplanar, tolerances in the probing apparatus make it impossible to ensure that the plane of the connection pads is parallel to the plane of the contact bumps, and in order to accommodate these tolerances, it is necessary to displace the DUT by 75 .mu.m in order to ensure contact at all connection pads. If the dimensions of the finger were changed to accommodate a displacement of 70-80 .mu.m (75 .mu.m.+-.5 .mu.m), the probe would become much less robust. If the probe were supported at a location further back from the root of the fingers, such that most of the deflection would be carried by the substrate rather than the fingers, the ability of the fingers to conform would be limited to 0.13 .mu.m/gram deflection produced at the fingers themselves.
The connection pads of the DUT are not coplanar, nor are the connection bumps on the probe. Assuming that the nominal plane of the connection pads (the plane for which the sum of the squares of the distances of the pads from the plane is a minimum) is parallel with the nominal plane of the contact bumps, the variation in distance between the connection pad and the corresponding contact bump is up to 5 .mu.m if both the DUT and the probe are of good quality.
At present, the connection points on an integrated circuit chip are at a pitch of at least 150 .mu.m, but it is expected that it will be feasible for the pitch to be reduced to about 100 .mu.m within a few years.
As the need arises to make connection at ever finer pitches, the stress in a probe of the kind shown in U.S. Pat. No. 5,221,895 increases. If the connection pads are at a spacing of 75 .mu.m, this implies that the width of the fingers must be less than about 50 .mu.m, and in order to keep the stress below the yield point, the height of the fingers must be at least 400 .mu.m.
The necessary height of the fingers can be reduced by employing a metal of which the yield point is higher than that of beryllium copper. For example, if the substrate is made of stainless steel, having an elastic modulus of 207.times.10.sup.9 N/m.sup.2, the maximum height of the fingers can be reduced to about 350 .mu.m. However, it follows that the deflection is reduced below that necessary to comply with typical height variations found in the industry. Additionally, the resistivity of stainless steel is substantially higher than that of beryllium copper, and this limits the frequency of the signals that can be propagated by the microstrip transmission lines without unacceptable degradation.
Although the microstrip transmission line has adequate characteristics for signals up to a frequency of 5 GHz, the so-called stripline configuration is desirable for higher frequencies.